Hobby CNC, the rotor way

The Something Awful Forums > Discussion > Hobbies, Crafts, & Houses > Hobby CNC, the rotor way

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Aurium posted:

Yes and no.

For hobby stuff a realtime patch for linux, or some of the better windows solutions, are easily good enough. Low end hardware, such as an arduino, effective performance is really no better or worse, though it does have more I/O and the possibility of measuring analog signals. The trade off is that you now need some kind of infrastructure to get the g-code to the uC rather than it all just being on the computer. You also don't need to find a parallel port. So at the low end it's really more about I/O and how you want to do it.

The finer your work the higher the step rate, providing you want to keep speed. The faster the step rate the more issues you have with mistimed pulses from os jitter. Real systems can have 100k-10m step rates.

The opposite side is the very low end hardware, such as an arduino, just isn't fast enough to produce steps at this rate either. At 100k step rate and 3 axis, you'd be talking 1 in every 26 instructions being a step pulse.* Now consider that you're working with very specific timed intervals, and might need to do math for acceleration.

Of course, uC's grow extremely fast in terms of clock speed without sacrificing determinism, and they can also have very high gpio counts. That's one solution, and another typical one is a FPGA pci (or sometimes the newfangled pcie) card, with the potential to offload work onto it too.

*16m instructions per sec / 100khz /3 axis / 2 instructions needed to make a pulse (high low)**

**I'm assuming can do useful work between the high low instead of wasting instructions with a delay, you don't have many left.

I just wanted to chime in on this a bit. I took a bare arduino and hooked up a A4988 stepper controller to it with three wires (step, dir, enable).

In AVR c, my loop was this:

code:

main()
{
write(PIN,LOW);
write(PIN,HIGH);
}

and the stepper spun along fairly well.
I then looked at the assembler and did the same thing. My program was only 40 bytes or so, and the motor spun MUCH, MUCH faster. Now, atmel is doing clever things with their 328s, so 1 clock=1 instruction. Other microcontrollers have execution speed that's a fraction of clock speed (1/4 usually).

Now clearly, my firmware isn't going to be able to control things as fast as the tightest-coded loop available. The best I've been able to get without missing steps is something on the order of 5000 steps/sec, which is 20RPM. If I enable servomotor control, it gets slightly worse, as the servo library is interrupt-driven. The servo jitters and there are annoying pulses in the movement of the axes. SDcard and LCD also on? Servo is even more jittery, and I start to lose top-end speed; only being able to get 3000 steps/sec.

I'm not sure what the answer to this problem is, but I'm sure it's going to be more, faster chips with less processing burden per chip, with well-designed code to get timings PERFECT.

# ¿ Oct 17, 2014 13:26

Adbot: ADBOT LOVES YOU

# ¿ May 5, 2024 16:58

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Aurium posted:

Have you tried any kind of acceleration ramping?

In marlin (one of the reprap arduino mega 2560 firmwares) step generation and timing is done by interrupt. It was kind of hard to find numbers quickly, but this guy's analysis indicates that he's getting 10-20k steps per second on his machine.

I'm on a Mega 1280; that may be the issue. I'll investigate whether it's worth it to move up to a 2560.

edit: I'm also using scavenged steppers. They say 12VDC 1.7A/phase. They're six-lead motors, and I've got them hooked up with only 4 leads, using the full windings. Is this the correct way to do things? I am now fundamentally confused as to the difference between unipolar and bipolar stepper motors.

double edit: A computer store nearby was having a yard sale and I ended up with a 2560 and LCD/button shield. Score. We'll see if this thing gets better.

babyeatingpsychopath fucked around with this message at 20:48 on Oct 17, 2014

# ¿ Oct 17, 2014 18:40

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Cockmaster posted:

I notice the Smoothieboard uses the same user interface software as the ATMega-based 3D pritner controllers.

Is there any such host software which would be effective for serious machining (say, with a Sherline lathe or mill)? All examples I've seen of using this stuff for CNC machines involve Dremel tools bolted to 3D printer frames. Could any of this software handle things like changing tools, or using a probe and toolsetter?

I just did some probing with my marlin firmware.

It supports up to 4 tools natively, and the code is fairly straightforward if you need more. If you've got a bigger arduino board (2560 in my case) you can easly support 64 tools, each with their own homing offset. You're limited at the moment to one z-probe sled for homing; which is what I assume your "probe and toolsetter" is. Right now, I have my brass hot end with an alligator clip on it and my aluminum print bed with another clip. That's my Z end stop. I know to within my limits of precision what 0.000 Z is. And since it homes at very low feed rates, my repeatability is better than .002, which I THINK is my z-axis precision anyway.

The gcode "T0" switches to tool0 already. Sprinter ignores that gcode if you don't have multiple extruders enabled, but marlin just accepts it and says "extruder 0".

It looks like Marlin is a really robust firmware. I'm seeing #defines for deltas and scadas; there should be no reason not to be able to fairly simply implement a 3-5 axis mill into this firmware, with a multitude of tools.

babyeatingpsychopath fucked around with this message at 03:25 on Oct 19, 2014

# ¿ Oct 19, 2014 03:14

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

unpacked robinhood posted:

I figure this is the closest to ideal thread to ask this:

What would be a suitable material to make custom light indicators, like those of a car dashboard ?

My idea was to draw my icons in whatever format a laser cutter will accept and have them done at the local hackerspace. I admittedly have 0 experience with laser cutting.
Is this workable at all ?

e: I would cut the wanted shape into the material and light it from behind through a translucent surface.

If your makerspace has a vinyl cutter, use that. Cut the vinyl and stick it to your clear stuff, then peel your designs away. You can pick your vinyl color and "clear" color separately, too.

# ¿ Jan 24, 2015 20:55

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Rakins posted:

I'm trying to figure out how to power and wire my 1.5kw 110v spindle as most are 220v. I'm looking for the cord the connects to the top of the spindle, I can't really find the right plug. I might just solder but it's a lot of power so I don't wanna.

Any markings on the plug? Looks a lot like an amphenol 4-pin industrial plug (97 series?), but a LOT of those look very, very similar.

# ¿ Jun 2, 2015 17:24

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Parts Kit posted:

Yeah, the plan would be to cut out spots for at least two fans, which I'm concerned will be a huge pain.

edit: Thought this was interesting, #2 recyclables are HDPE and can be melted and formed into blocks at home pretty easily.
https://www.youtube.com/watch?v=kUR6_bQLU-E

Ammo cans are thin metal. Any normal hole saw works in them just fine.

# ¿ Dec 13, 2015 23:22

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Ambihelical Hexnut posted:

Cool. I�m not familiar with the 3018 but the problems you figure out on any machine that make the difference between enjoying it and hating it are:

Workholding reliably and easily making toolpaths that don�t interact with your clamps. I use painters tape and superglue.
Tool changing quickly and easily and adjusting your offset without a hassle. I jog Z up, drop the old tool, put the new one in all the way, jog down to z+6mm, loosen the collet until it drops to stock, tighten the collet, then redefine current z as zero. Touch probe routines and fixed offset tool holders are better solutions.
Repeatable xyz positioning by limit/homing switches.
Keeping dust down. I have a magnetic dust boot with a collector hose hung above the machine.
Managing and cutting stock, and holding the trash from scraps. This really is a pain in the rear end, get ready to dispose of a shitload of plywood scraps. And your work space is never big enough when you need to cut down 4x8 sheets so they fit on the 3x3 router.

I got one of those 3018s and I'm glad this thread is still active.

They run GRBL by default, and I'm sending commands with Candle (grblcontrol). I come from a 3d printer background with gcode where everything is done in positive space, and milling machines and routers home to the far ends so the "machine space" is all negative. I'm cool with this.

I'm trying to figure out how to use the various other coordinate thingies effectively. I know G53 is machine coordinates, and they're all negative, and if I ever go above 0 on a G53 move, then I hit an endstop. But how about G54-G59? I have a probe available to get tool offsets, so it seems like this is a thing, right?

How do machinists move into different workspace coordinate systems in gcode? Is there a website I can read? I don't want to watch a half-hour youtube just to get the twenty seconds of button mashing I need; I'd prefer to read some complete pedant's explanation for half an hour instead.

# ¿ May 25, 2019 08:18

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Methylethylaldehyde posted:

https://www.youtube.com/watch?v=user?krayvis?videos does a pretty good job of explaining how to work poo poo on the grbl style machines.

Link didn't work, but I found the guy and watched some of his videos.

So, lemme see if I have this right.

I make some toolpath that carves a 5mm by 5mm dickbutt on something. I then home my machine. It's now at g53 0,0,0, and my workpiece is way off in the negatives. I jog over to the piece and run G54, then a probe operation. I'm now at G54 0,0,0. If I run my toolpath, it'll carve right there. I could have also set my probe on g55 or whatever and run the same path? How does this relate to tool offsets?

Or, to use a more complicated example, I have my 5mm dickbutt toolpath. I probe at a (G54) corner, then move 10,10, then say g55=0,0 and run, then g54 to 20,10, then g55=0,0 and run again?

That's pretty slick.

babyeatingpsychopath fucked around with this message at 03:32 on May 30, 2019

# ¿ May 30, 2019 03:28

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

shovelbum posted:

I am looking at CNC wood routers and this thread came up, are any of the smaller home machines like the 3018 good for this? I am mostly looking for something that can accurately carve small enclosures and face plates (think 4x4x1.5" max) from wood or plastic. I am guessing the optimal way to do it is a small router/table and jigs though.

I've got a 3018 and it works a treat for this. I got the 1/8" collet and use Dremel bits. It's a bit fiddly to get trammed in and square, after that, it produces tons of sawdust no problem. Also engraves aluminum, brass, and other soft metal, and can mill PCBs.

# ¿ Jul 30, 2019 02:56

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

I may be getting a big-boy CNC machine. Our CNC vendor has decided to be a bunch of asses, so my boss told me to find a CNC table.

I'm cutting stuff in (max) 1/8" aluminum. Could this be a DIY thing, or do we need to go find an actual CNC router? What kind of spindle power do we need to make full-depth cuts on 1/8" aluminum?

# ¿ Oct 10, 2019 00:21

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Call it 24"x24", +-.003. Smallest typical feature .063" holes. If i could spot those at least, I'll drill them with appropriate drill bits.

# ¿ Oct 10, 2019 04:56

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

After slowly upgrading my 3018 engraver to run a laser on a 600x600mm carriage and never getting the frame rigidity good enough to keep y-backlash down....

I bought a Shapeoko XL. Now I'm going to cut aluminum parts to stiffen up the laser

Each individual part on this weighs more than the entire 6060 (minus spillboard). Like, the Z-axis is heavier, Y1 axis, Y2 axis, X-gantry aluminum extrusion....

One thing I didn't realize is the Shapeoko doesn't have spindle control; that's 100% manual. I can work with that, it was just a bit of a surprise.

# ¿ Oct 3, 2020 14:38

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

I got the trial of GWizard for a feeds and speeds calculator.

Does 26000 RPM @ 3000/min sound right for a 1/8" single-flute cutter in HDPE/delrin? It seems REALLY fast. It's telling me 4mm max depth of cut with 1.27mm stepover; 550 mm/min helical ramp in. Finish pass is 1200mm/min .5mm stepover.

# ¿ Oct 5, 2020 12:23

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

sharkytm posted:

Yup. Delrin can be machined crazy fast. 100+ ipm and high spindle speed is how my machinist does it. Chip control is a pain on the lathe for sure, not so bad on a mill.

Spindle speed with affect finish, so monkey with that to get an acceptable product. Check your outside corners too, high speed can chip them.

Ok, so the cuts are looking great at low DOC, but my calculator says I can do like 4mm at a time. When I go 4mm (at same stepover) I get melting and the surface finish is garbage. Which direction do I go? Slower feed? Slower spindle? More cooling? I've got a vacuum pointed at it for chip clearing, but I can turn on air blast if I need to.

# ¿ Oct 5, 2020 17:13

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Sagebrush posted:

If the material is melting you obviously need more cooling. For plastics, soapy water works very well.

If you can't increase the cooling, then you decrease the heat by reducing energy input (spindle speed) or friction (depth of cut).

The shapeoko forums are telling me the correct action is "increase chip load" and get my MRR up to 40mm³. That is, amm, agressive.

# ¿ Oct 5, 2020 19:07

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Well, I turned the feed rate up 50% (so now moving 6000mm/min on a full-depth slot with a 1/8" plastic-cutting bit) with 80PSI air blast. The chips are coming out and nothing's sticking to the bit, but surface finish is still all gummy at the top.

# ¿ Oct 5, 2020 21:14

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Ok, I cut some test zig-zag patterns at different feeds. All 1/8" depth (3.175mm, 1 tool diameter). I started at 4000mm/min and went all the way up to the machine's maximum rate of 10,000mm/min, increments of 200mm/min. Spindle RPM was set to 18,500 as recommended on the shapeoko forums. The sweet spot for cutting at that depth seems to be 7200mm/min. There's just the tiniest burr at sharp corners. Below~6200, I get long, stringy chips. Below about 5000 or so, stuff is sticking to the bit. Above 8600, the surface finish is garbage. The machine skips steps in X at 9800mm/min.

Vacuum only, no air blast. Brand new single-flute up-cut bit.

I'm going to try for some pockets now, instead of slots. Plastic is crazy.

# ¿ Oct 5, 2020 23:50

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Methylethylaldehyde posted:

The calc gives you 'this is what an ideal, six-figure machine could do with perfect workholding', based on the material and the endmill selected. Most of them don't do any de-rating for using a substantially smaller and less rigid machine.

On a lighter machine you want less width of cut and more depth of cut as a rule, using HSM tool paths preferably. 15-25% width of cut, 2x diameter depth of cut is a pretty normal recipe. Slotting is basically the worst possible thing to do in plastics, since they tend to remelt and make a huge mess more or less instantly, you want lots of long, thin chips carrying the heat away from the part, and slotting really doesn't do that, unfortunately.

Once you DO find a recipe you're happy with for each type of cut (profiling, slotting, pocketing), write it down somewhere and keep it handy, finding the recipe that makes your machine happy is like 80% of the work in CNC machining, the other 20% are figuring out how to hold the part.

When you say 15-25% WOC, do you mean 15-25% tool diameter?

What is a good workflow to start to determine a "recipe" for cutting? How big of a feature should I go for? Like, should I be cutting a rectangular path of X tool diameters using more and more stepover to find a good setting? What about holes? When I'm looking at feeds and speeds in Fusion 360, there's a bunch of stuff I can set and it'll adjust everything else for me. When cutting plastic, what number should be the pivot for everything else? Do I want to keep "feed per revolution" or "feed per tooth" constant, or is it something else? Again in Fusion, I assume the "2d adaptive" is the HSM operation. If not, what ops should I be going for for my small machine?

The calculator is GWizard, and it touts itself as having specific compensations in for small machines; the writer of the software got a shapeoko specifically to check. I have "safe mode" turned on, which says it makes stuff more conservative specifically for the machine selected.

I'm frustrated because almost all of the data out there is "this is what works for me in my machine" and not "from first principles, here's how I determined what settings I can use based on an arbitrary cutter and material"

# ¿ Oct 6, 2020 13:44

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

sharkytm posted:

Cool, glad you got something working. I'd be a little suspicious of those cutters, and you might do better with a YG-1 two flute AL geometry cutter. Or a new HSS cutter, TBH. Be glad you're not cutting nylon, PTFE, or polyethylene. Ugh. Or zinc or lead or copper.

I will be cutting nylon, PTFE, and PE. Please impart your wisdom!

edit:

I would like to sincerely thank sharkytm for helping offline with my HDPE problem.

The solution was less DOC, less stepover, less RPM. I ended up with:
Single-flute upcut endmill, 3.175mm diameter. .5D (1.58mm) DOC, 40% (1.27mm), 11000 RPM on the spindle, 4000mm/min feed. .22mm chip load, 2cc MRR, 180m/min surface speed.
That's a hair over 150IPM. ~.1" chipload, 1.25 cu in MRR, 360SFM.

I also now know how to turn what the calculator spits out into what fusion likes and what that means on the machine.

Learning machining from first principles is great!

babyeatingpsychopath fucked around with this message at 23:21 on Oct 6, 2020

# ¿ Oct 6, 2020 13:48

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Methylethylaldehyde posted:

Also, the power of Google compels you! Here is a speed and feed chart for the Shapeoko 3 for most of the materials you'd actually want to cut with it. You should be able to use the rough ratios between the materials to get something that would work.

Good info. As to that document, I have it printed and in front of me, but it's for a 3-flute 1/4" endmill. What columns do I extract from something like that if I have a 2-flute 1/8" or 1-flute 1/8" or whatever?

My general usage case for this machine is 95% pocketing, 5% part cut out of a sheet. If slotting is bad for part cutout, should I just design in a 2xD gutter around the part and pocket that, then do a finishing cutout pass?

Last thing, which is a big limitation, I know, but the "spindle" in a shapeoko is a pocket router. It's got six discrete steps from 11,000 to 31,000 RPM. I'm seeing that "lowest possible RPM and maximum possible feed" is the way to run with plastic. It looks like other people are having success with the machine as-is; I know that cleverness and computers can make this work.

# ¿ Oct 7, 2020 18:21

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Methylethylaldehyde posted:

Tabs are your friend for cutting parts out. You can do a 2xD cutout, leave ~10 thou for a finish pass, and have a few little tabs holding the part that you can just cut off with flush cut snips. There is an option in fusion360 to add them to the part. Here is a video on it.

That document just gives you rules of thumb for how relatively obnoxious a material will be. Compare the HDPE and the Nylon recommendations, and you'll see that while the feed is more or less the same, the depth of cut goes down and the speed drops about 25%. Since you have a good recipe for HDPE, just dropping the depth of cut some should get you in the right ballpark for nylon.

Well, I went to the local plastic supplier today and got 4x8 (48 x 96 inches) sheets of: cast acrylic, extruded acrylic, hdpe, uhmw, polycarbonate, abs, pvc, and polyethylene. Time to cut some coupons.

# ¿ Oct 8, 2020 00:55

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Trebuchet King posted:

i hate to tell you this bud but that's a big ol 4 0 4

on the subject of how many passes to do, i work in a sign shop and we make some pretty neat things, but if time permitted i would always, always, always do more passes. deadlines mean i frequently gotta cut clear through, though.

Just ran the acrylic today. It machines like a dream compared to HDPE. The table said d/4 for DOC and 80% feedrate of HDPE. Plugged it in, and bam. Amazing cuts. I got my feed off on a full-depth slotting and got some melting on the bit, but it didn't affect the finish bad enough to worry about.

Then I threw it in the annealing oven, and apparently I had extruded acrylic, not cast, and all the bubbles came out and the thing just melted into a semblance of its former self.

Learning.

# ¿ Oct 16, 2020 07:34

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Martytoof posted:

That's not really out of the ordinary for Mac apps so I'm not terribly surprised, just annoyed

Another Fusion question. I'm designing a cookie cutter, the base of which is an SVG I'm importing as a sketch. I want to create the outer part that slices the dough to size so I offset the inner design by 5mm to be the inner wall of the "cutter", then I want to create the outer wall:

1. If I try to offset the offset, that is apparently something I'm not allowed to do.
2. If I offset the outside of the design to create the outer edge, that works but then .. and apologies, I don't know the Fusion term for what I'm building here .. the two offsets are just their own thing and don't form any kind of object that I can then extrude up.

Super crude sketch of what I'm trying to do:

e: Oh sorry I forgot to ask an actual question. Given an SVG sketch with a discrete outer "edge" that I can offset off, how would I go about building an outside "shell" that I can then extrude up which is offset by 5mm from the sketch itself, which follows the contours of the sketch? I know my little doodle above is just a circle but the SVG itself is more complex.

e2: Wait, no, this time it worked? The two offset edges created an extrudable profile. I swear it wasn't working earlier. Hmm, ok maybe PEBCAK.

So the way I'd do this in fusion is a set of multiple extrudes. Extrude EVERYTHING up to get a big round disc. Then your first one is extruded up from the base of your disc (mode:cut), then your shape extruded down from the center of the cutout (mode:join).

When doing SVGs, I've found it easiest the saved for "stroke as path" so you have all outlines to extrude, instead of just raw circles. You're left with a bunch of stair-step (wedding-cake) lines in your extrudes, which you can then throw draft/bevel/chamfer on to get edges.

# ¿ Oct 22, 2020 17:12

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

https://imgur.com/a/QZe2Nul

I hope this gallery comes through.

No idea where pic5 got off to on my hard drive; that was the money shot. You use the "draft" feature, select the base plane of the feature you extruded up in step 4, then select on a single closed set of curves, then draft that to make a point. symmetric draft works well, ~10deg or so, but adjust in 1deg increments from 5deg. If you get too much draft or your spline is too curvy, then fusion will chew for a while and then say "I give up, this made your surface non-manifold."

Going from pic4 to pic6 took four drafts. one for the exterior circle, then one for the mouth and each eye. I didn't try too too hard to make a reasonable .svg to start with, either. just 20 seconds in inkscape to throw some nontrivial curves up.

# ¿ Oct 23, 2020 07:45

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Dominoes posted:

I've made some changes since, like moving the ports shown on the left to the back, and spacing them out, since even the drill press can't do holes on the long side, and I need to account for grommets in the spacing.

It's a poor craftsman that blames his tools. You need to fixture the box in the drill press somehow. Even throwing a chunk of wood on the table, turning it out of the way, and then backstopping your "long edge" so you're drilling from the plastic into the wood is better than just free-air drilling.

I've made dozens of little enclosures in Hammond boxes in aluminum, ABS, PVC, and whatever garbo acrylic they use using nothing but a drill press, hand drill, and hand files. It may take a smidge longer, but unless you're a hamfisted lummox with no attention to detail, no patience, and a cavalier attitude, it is absolutely possible to get professional-looking boxes in plastic by hand.

# ¿ Oct 27, 2020 22:33

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

sharkytm posted:

The other trick is to design the pattern in CAD and print it out on sticker paper. Then you've got nice clean lines that show you where to cut/drill/file, and it's easily removed with some heat or iso alcohol. A drill, nibbler, hand files, and maybe some sandpaper is all you really need. However, as soon as you're talking about more than 10-20 (depending on complexity), just have Hammond/Budd/ProtoCase/PolyCase make them. I just ran some numbers through PolyCase's instant quote system, and it's super cheap.
A DC-59F enclosure: https://www.polycase.com/dc-59f which is like $12.50 to start
with 4 sides of the base needing machining, with 14! total features, plus the lid having 4 features came out to:
QTY:1 (stupid money, $238 each)
QTY:5 $61 each (still spendy)
QTY:10 $38 each (getting there)
QTY:25 $25 each (very affordable)
QTY:50 $19 each (Now they're cheap)

Figure out your time per unit, assume you'll gently caress up some, and it gets very affordable, very quickly. There's a $100 setup fee, but there are other companies. Compare and find out.

Agreed. There are a bunch of good tips. I lay my stuff out in CAD, print it at 100% scale, verify measurements, then use glue stick or 3M 77+ to stick the stuff to the enclosure. Use a razor knife to cut on your lines, center punch your holes, drill undersize, file to size.

# ¿ Oct 27, 2020 23:25

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

I want to test my V-carving-with-cutout paths on a few different materials.

I have 1/8, 1/4, 1/2" MDF, 1/4, 1/2, 3/4" plywood, 3/4" OSB, and some various thicknesses of solid wood.

What I want is to cut out the same model that I've already designed in Fusion 360, and just have some CAM parameter magic to spit out like ten different files for each of the various thicknesses/speeds. Is there a way to parameterize CAM? I am not looking forward to setting up a model for each thickness of material, then a NC program for each combo by hand.

# ¿ Nov 5, 2020 01:48

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

I ended up in the "manufacture" space for all of this. I have one model at nominal dimensions (1/4").

I then generated a setup for "relative size box" with side offset 3/16" and top and bottom offsets of 0". I then did all my paths with heights based on "stock top" or "stock bottom." I then duplicated this setup, changed the "bottom offset" in the setup to 1/4" and now the paths are all for 1/2" material. Duplicate again, -1/8" offset, and the paths are for 1/8" material. I used adaptive processes for pretty much everything, so I set my feedrates to 1111, 2222, 3333, 4444 etc. I then wrote a little python script to take "1/4" raw" and spit out "1/4" plywood" and "1/4 MDF" and whatnot by just replacing all the feedrates with stuff out of a table.

Not as elegant as having a bunch of parameters, and if this gets into dozens of ops (instead of 1 engrave, 2 holes, and a cutout) this may get unweildy.

# ¿ Nov 6, 2020 14:55

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

All of this assumes that's a 3018 pro. If so, then your available range of spindle speeds are 700-1000, referenced to a max of 1000. These numbers have no relation to actual RPM unless you've measured it.

These things have a "10k rpm" spindle by default. At 24V and no load, I measured mine at 10760 RPM. Not too shabby. RPM scaled linearly at no load down to 50%, then not so much linearly down to 30%, where it stopped self-starting.

That said, I'd run it to 100% RPM, plunge 2mm into a piece of MDF at feedrate 100(mm/min), then use the jog to drive 30mm (still feed 100). At 100% (M3S1000) I was getting 9700RPM. At S800, I was getting 7600RPM. That's not 80%, but close. At S700, it was 5100RPM. Below that, it was straight garbage.

You need to do something like that to test what actual feed you can run. Keep your spindle above 70%, 100% preferable. Plunge down, then use the jog keys to just drive around in the plastic at various feedrates and find something that works at all.

# ¿ Nov 12, 2020 02:42

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Mofabio posted:

I would buy the heaviest lathe you can realistically move with yourself and a friend, and convert that to CNC. The 8x16 at 150lbs, or even better, the 9x19 at 300lbs, which has a real DIY community. Or old iron. On mine I use a 200mm THK SKR33 for X, 300mm SKR33 for Z, Gecko G540, 2x NEMA23s, DIY apron, motor pulley, and RPM/index sensor, an old 4-jaw that's too big (but holds work true, unlike grizzly's 4-jaws), and a 240V 3-phase motor on VFD, running thru RTAI linuxcnc. If I did it again, I would get an ethernet mesa card like the 7I96, an ODROID H2+ on linuxcnc preempt-RT, and a smaller 4-jaw from Bison. And I would've paid more attention in lathe class. Good luck.

I know what you're saying here, but could you break it down a little bit for those who aren't armpit-deep in controllers and servos and slides?

# ¿ Nov 16, 2020 16:23

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Anyone happen to have flashed grbl onto a melzi board before? I'm trying to get pin mappings going because I smoked one channel of my laser's board.

# ¿ Nov 20, 2020 21:10

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

babyeatingpsychopath posted:

Anyone happen to have flashed grbl onto a melzi board before? I'm trying to get pin mappings going because I smoked one channel of my laser's board.

Just to follow up. GRBL won't compile onto a melzi because the atmega 1284p doesn't have enough hardware PWM channels. Marlin is the same way, just because of the way it's written.

I broke out my RAMPS board (last updated 25 Oct 2014, with marlin 1.0.0 firmware) and _seem_ to have the board running on the latest Marlin 2.0 firmware and driving steppers. I'll get it connected to the machine tomorrow.

Marlin is already a little bit more irritating than GRBL, since it doesn't seem to do modal gcode states in the linuxcnc/grbl/mach3 way. That is, you type G1 F800, then all your next moves are in g1, with a feedrate of 800. but you can't type x1, y1, x-1, y-1. Marlin wants the G1 every. single. line.

If there's not a Marlin post for Fusion then I guess I'm buying a replacement GRBL board.

# ¿ Nov 22, 2020 02:54

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Triple post because that's how I roll:

Marlin for RAMPS worked great, but it didn't play nice with any of my control software, so I found GRBL for RAMPS. There was a fork of GRBL 0.9 for the sanguinolulu boards (of which melzi is a shadow/offshoot) but the pin numbers didn't line up and getting the code size right with the limited hardware looked more trouble than it's worth.

So I've got full GRBL 1.1 on RAMPS, and I'm pretty happy with it. RAMPS has so many more pins I'm flush with them, which is nice. I have separate inputs for max and min endstops, plus the ability to have a dedicated probe for Z, plus a toolsetter. This board (from 2014) will happily drive 5 axes, and the ATMega2560 seems to have the horsepower for it.

I'm probably going to run through the GRBL code and update some of the reporting features to make it a bit nicer for my use, but the code is very tight, where they're using bit-packing and 8-bit values to store 7 pieces of information in a spot where I really would like to have 18 pieces of information. Refactor ahoy.

Anyway, off to the machine to see if I can't burn some stuff.

# ¿ Nov 23, 2020 14:20

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

honda whisperer posted:

If you do cnc things with coolant and can't see buy an oleophobic screen protector. I got a pair for an ipad for $10.

Camera focused on the droplets but in person it's night and day better.

Interesting. I wonder if you could use those ceramic coatings designed for headlights that "REPEL WATER, OIL, BUGS, DIRT, GRIME!" to any effect.

# ¿ Dec 13, 2020 18:36

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

NewFatMike posted:

Anyone have a preferred chipload/feeds and speeds calculator? I was eyeing FSWizard Pro, but it looks like its recommendations are based on having a machine that's completely cast iron and the size of a schoolbus.

I can work with that, but even a calculator aiming at "light duty" machines like a Tormach would be a delight. Taig folks have a favorite?

I got GWizard on sale and it gives great results on my shapeoko. If I throw its "modified chip load per tooth" into fusion 360, then fusion just auto-calculates everything (a bit conservitavely) and it comes out great. So far it's worked really well in light aluminum, all the various plastic flavors I've put it through, and hard/softwood.

# ¿ Dec 30, 2020 08:36

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

BattleMaster posted:

So I finally got the parts in and have my 3018 Pro router hooked up to LinuxCNC via parallel port. Well, just the steppers so far; I still need to build a circuit for the spindle.

I have an issue where AXIS is recording half the real distance. As in, if I move the X-axis of the router completely from 0 to 300 mm, Axis thinks it moved only 150 mm.

I checked the thread pitch several times, the microstepper switches on the stepper driver several times, and the only thing that seems to make sense is that these motors actually only have 100 steps per revolution instead of 200. If I enter 100 steps per revolution into stepconf, AXIS shows the distances correctly.

So is this a thing? I know these things are built to a very low cost, but are there really extra-lovely 100 steps per revolution motors?

That, or your controller has a higher microstepping than you expect. Or your 'thread pitch' isn't actual pitch and you've got more or fewer starts on your screws so your effective pitch is different.

Fix whatever variable makes the machine and the software report correctly, and don't worry about it too long.

babyeatingpsychopath fucked around with this message at 02:42 on Jan 14, 2021

# ¿ Jan 14, 2021 02:38

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Dominoes posted:

Thoughts on this approach? The end result is a separate script that strings together the results of these, and outputs to a file. Going to do a helical-entry circle-cutter next.

Are you making a gcode postprocessor from first principles for a reason? Why aren't you making shapes in CAD and then running your software's CAM option and then just hitting "go" on the machine?

edit: edit edit: moved.

babyeatingpsychopath fucked around with this message at 17:00 on Feb 9, 2021

# ¿ Feb 9, 2021 16:50

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Dominoes posted:

More like a g-code generator. I don't have a CAD model, and am bad at CAD software. (Was messing with FreeCad earlier this week for a diff proj, but have a long way to go) Maybe that's the answer. Hand-written G-Code seems to be a unsustainable.

Ok, well. CAM without CAD is really, really, really hard. Like, trying to program a microcontroller in assembly with the datasheet. CAD -> CAM tools exist for the same reason compilers exist.

If you're just trying to do a linear ramp from point A to point B, then a G01 move will do that. If you look at a big facing cut as output by Fusion or something, it'll have

code:

(RAMP )
g00 x0y0z0
g01 Feed
x1y1z1
(END RAMP)
(FACE)
x2y2
x3y3
x4y4
...
(END FACE)

# ¿ Feb 9, 2021 17:00

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Dominoes posted:

OpenScad looks fun on its own! Diving in. I'm suspicious it'll be more appealing than the GUI-based ones, where I frequently fight to find to click the right faces. The mouse feels like a clumsy instrument for navigating 3D space in the way FreeCAD and Fusion operate.

I have a hunch a G-code generating script might be the most elegant solution to my specific use-case here: Making a few linear and circular cuts on the face of an enclosure. I still plan on learning FreeCad's path tool for more-complicated future projects, and in case it does prove better for this case.

Ie, the resulting file will look something like this:
Python code:
    cfg = machine.Config(
        Coord(0.0, 0.0, 10.0),
        5.0,
        100.0,
        100.0,
        1.0,
    )

    steps = cfg.setup()

    steps.extend(cfg.ramp_line(Coord(0, -6, 0), Coord(0, 6, 0), 10.0, -3.0))
    steps.extend(cfg.ramp_circle(Coord(5, 5, 0), 3.0, -3.0))
    steps.extend(cfg.ramp_circle(Coord(5, 10, 0), 3.0, -3.0))
    steps.extend(cfg.move_to_idle())

    machine.make_file(steps, "water_mon_top")

If that's the case, just look at how your machine interprets G0, G1, G2, G3 codes, some M-codes, and then just write the g-code by hand.

Like, the minimum program you have is 10 lines of gcode, and you've managed to condense it into 20 lines of python.

# ¿ Feb 9, 2021 19:54

Adbot: ADBOT LOVES YOU

# ¿ May 5, 2024 16:58

babyeatingpsychopath: Oct 28, 2000; Forum Veteran

Ambrose Burnside posted:

a lot of the discussion about openscad i've found online comes off as completely out to lunch and out of touch with the existing state of 'real' CAD as it exists in industry; i understand it's partially because the people who like it have verrrry different backgrounds, proficiencies and comfort areas than i do, but i don't even like calling it 'CAD' because it's all polygons with no NURBS representations. like, so far as i know true circles don't exist in openscad, just polygon approximations; that's useless to me as a designer. i need circles that satisfy the mathematical conditions of a circle, not straight-line-based approximations. here's two good posts from a programming-informed perspective about openscad from the 3d printing thread, the first a response to me asking why anybody uses openSCAD, the second critiquing it as a CAD platform:

nowadays i just don't see the justification for doing much of anything with openSCAD unless you're specifically developing free/open-source parametric parts where accessibility is critical, but you can always do those things *better* via any number of CAD alternatives

I agree with this entire post and use OpenSCAD a lot for making little doohickeys for the 3d printer.

If it's gotta have tolerance or geometries more complex than cubes and stuff, then off to Fusion I go.

edit: one final complaint. When previewing, openSCAD takes advantage of any graphics hardware you've got, and appears to be multithreaded. When rendering, it's multithreaded, but doesn't take advantage of your GPU. I know for a fact my GPU can render a STUPID number of polygons, intersect them, and produce a solid geometry. Preview of some stress-test designs takes ~a second. Rendering can be over 20 minutes.

babyeatingpsychopath fucked around with this message at 15:37 on Feb 11, 2021

# ¿ Feb 11, 2021 15:01

The Something Awful Forums > Discussion > Hobbies, Crafts, & Houses > Hobby CNC, the rotor way

«‹›2 »